def pretrain(self, dataset: BaseADDataset, optimizer_name: str = 'adam', lr: float = 0.001, n_epochs: int = 100,
lr_milestones: tuple = (), batch_size: int = 128, weight_decay: float = 1e-6, device: str = 'cuda',
n_jobs_dataloader: int = 0):
"""Pretrains the weights for the Deep SVDD network \phi via autoencoder."""
self.ae_net = build_autoencoder(self.net_name)
self.ae_optimizer_name = optimizer_name
self.ae_trainer = AETrainer(optimizer_name, lr=lr, n_epochs=n_epochs, lr_milestones=lr_milestones,
batch_size=batch_size, weight_decay=weight_decay, device=device,
n_jobs_dataloader=n_jobs_dataloader)
self.ae_net = self.ae_trainer.train(dataset, self.ae_net)
self.ae_trainer.test(dataset, self.ae_net)
self.init_network_weights_from_pretraining()
class DeepSVDD(object):
"""A class for the Deep SVDD method.
Attributes:
objective: A string specifying the objective.
net_name: A string indicating the name of the neural network to use.
net: The neural network \phi.
ae_net: The autoencoder network corresponding to \phi for network weights pretraining.
optimizer_name: A string indicating the optimizer to use for training the Deep SVDD network.
ae_trainer: AETrainer to train an autoencoder in pretraining.
ae_optimizer_name: A string indicating the optimizer to use for pretraining the autoencoder.
results: A dictionary to save the results.
"""
def __init__(self, objective: str = 'one-class', nu: float = 0.1):
"""Inits DeepSVDD with one of the two objectives and hyperparameter nu."""
assert objective in (
'one-class', 'soft-boundary'
), "Objective must be either 'one-class' or 'soft-boundary'."
self.objective = objective
assert (0 < nu) & (
nu <= 1), "For hyperparameter nu, it must hold: 0 < nu <= 1."
self.net_name = None
self.net = None # neural network
self.optimizer_name = None
self.ae_net = None # autoencoder network for pretraining
self.ae_trainer = None
self.ae_optimizer_name = None
self.results = {
'train_time': None,
'test_auc': None,
'test_time': None,
'test_scores': None,
}
def pretrain(self,
dataset: BaseADDataset,
optimizer_name: str = 'adam',
lr: float = 0.001,
n_epochs: int = 100,
lr_milestones: tuple = (),
batch_size: int = 128,
weight_decay: float = 1e-6,
device: str = 'cuda',
n_jobs_dataloader: int = 0):
self.ae_net = build_autoencoder(self.net_name)
self.ae_optimizer_name = optimizer_name
self.ae_trainer = AETrainer(optimizer_name,
lr=lr,
n_epochs=n_epochs,
lr_milestones=lr_milestones,
batch_size=batch_size,
weight_decay=weight_decay,
device=device,
n_jobs_dataloader=n_jobs_dataloader)
self.ae_net = self.ae_trainer.train(dataset, self.ae_net)
self.ae_trainer.test(dataset, self.ae_net)
def save_model(self, export_model, save_ae=True):
"""Save the model to export_model."""
#net_dict = self.net.state_dict()
ae_net_dict = self.ae_net.state_dict() if save_ae else None
torch.save({'ae_net_dict': ae_net_dict}, export_model)
def load_model(self, model_path, load_ae=False):
"""Load model from model_path."""
model_dict = torch.load(model_path, map_location='cpu')
if load_ae:
self.ae_net.load_state_dict(model_dict['ae_net_dict'])
def save_results(self, export_json):
"""Save results dict to a JSON-file."""
with open(export_json, 'w') as fp:
json.dump(self.results, fp)
class DeepSVDD(object):
"""A class for the Deep SVDD method.
Attributes:
objective: A string specifying the Deep SVDD objective (either 'one-class' or 'soft-boundary').
nu: Deep SVDD hyperparameter nu (must be 0 < nu <= 1).
R: Hypersphere radius R.
c: Hypersphere center c.
net_name: A string indicating the name of the neural network to use.
net: The neural network \phi.
ae_net: The autoencoder network corresponding to \phi for network weights pretraining.
trainer: DeepSVDDTrainer to train a Deep SVDD model.
optimizer_name: A string indicating the optimizer to use for training the Deep SVDD network.
ae_trainer: AETrainer to train an autoencoder in pretraining.
ae_optimizer_name: A string indicating the optimizer to use for pretraining the autoencoder.
results: A dictionary to save the results.
"""
def __init__(self, objective: str = 'one-class', nu: float = 0.1):
"""Inits DeepSVDD with one of the two objectives and hyperparameter nu."""
assert objective in ('one-class', 'soft-boundary'), "Objective must be either 'one-class' or 'soft-boundary'."
self.objective = objective
assert (0 < nu) & (nu <= 1), "For hyperparameter nu, it must hold: 0 < nu <= 1."
self.nu = nu
self.R = 0.0 # hypersphere radius R
self.c = None # hypersphere center c
self.net_name = None
self.net = None # neural network \phi
self.trainer = None
self.optimizer_name = None
self.ae_net = None # autoencoder network for pretraining
self.ae_trainer = None
self.ae_optimizer_name = None
self.results = {
'train_time': None,
'test_auc': None,
'test_time': None,
'test_scores': None,
}
def set_network(self, net_name):
"""Builds the neural network \phi."""
self.net_name = net_name
self.net = build_network(net_name)
def train(self, dataset: BaseADDataset, optimizer_name: str = 'adam', lr: float = 0.001, n_epochs: int = 50,
lr_milestones: tuple = (), batch_size: int = 128, weight_decay: float = 1e-6, device: str = 'cuda',
n_jobs_dataloader: int = 0):
"""Trains the Deep SVDD model on the training data."""
self.optimizer_name = optimizer_name
self.trainer = DeepSVDDTrainer(self.objective, self.R, self.c, self.nu, optimizer_name, lr=lr,
n_epochs=n_epochs, lr_milestones=lr_milestones, batch_size=batch_size,
weight_decay=weight_decay, device=device, n_jobs_dataloader=n_jobs_dataloader)
# Get the model
self.net = self.trainer.train(dataset, self.net)
self.R = float(self.trainer.R.cpu().data.numpy()) # get float
self.c = self.trainer.c.cpu().data.numpy().tolist() # get list
self.results['train_time'] = self.trainer.train_time
def test(self, dataset: BaseADDataset, device: str = 'cuda', n_jobs_dataloader: int = 0):
"""Tests the Deep SVDD model on the test data."""
if self.trainer is None:
self.trainer = DeepSVDDTrainer(self.objective, self.R, self.c, self.nu,
device=device, n_jobs_dataloader=n_jobs_dataloader)
self.trainer.test(dataset, self.net)
# Get results
self.results['test_auc'] = self.trainer.test_auc
self.results['test_time'] = self.trainer.test_time
self.results['test_scores'] = self.trainer.test_scores
def pretrain(self, dataset: BaseADDataset, optimizer_name: str = 'adam', lr: float = 0.001, n_epochs: int = 100,
lr_milestones: tuple = (), batch_size: int = 128, weight_decay: float = 1e-6, device: str = 'cuda',
n_jobs_dataloader: int = 0):
"""Pretrains the weights for the Deep SVDD network \phi via autoencoder."""
self.ae_net = build_autoencoder(self.net_name)
self.ae_optimizer_name = optimizer_name
self.ae_trainer = AETrainer(optimizer_name, lr=lr, n_epochs=n_epochs, lr_milestones=lr_milestones,
batch_size=batch_size, weight_decay=weight_decay, device=device,
n_jobs_dataloader=n_jobs_dataloader)
self.ae_net = self.ae_trainer.train(dataset, self.ae_net)
self.ae_trainer.test(dataset, self.ae_net)
self.init_network_weights_from_pretraining()
def init_network_weights_from_pretraining(self):
"""Initialize the Deep SVDD network weights from the encoder weights of the pretraining autoencoder."""
net_dict = self.net.state_dict()
ae_net_dict = self.ae_net.state_dict()
# Filter out decoder network keys
ae_net_dict = {k: v for k, v in ae_net_dict.items() if k in net_dict}
# Overwrite values in the existing state_dict
net_dict.update(ae_net_dict)
# Load the new state_dict
self.net.load_state_dict(net_dict)
def save_model(self, export_model, save_ae=True):
"""Save Deep SVDD model to export_model."""
net_dict = self.net.state_dict()
ae_net_dict = self.ae_net.state_dict() if save_ae else None
torch.save({'R': self.R,
'c': self.c,
'net_dict': net_dict,
'ae_net_dict': ae_net_dict}, export_model)
def load_model(self, model_path, load_ae=False):
"""Load Deep SVDD model from model_path."""
model_dict = torch.load(model_path)
self.R = model_dict['R']
self.c = model_dict['c']
self.net.load_state_dict(model_dict['net_dict'])
if load_ae:
if self.ae_net is None:
self.ae_net = build_autoencoder(self.net_name)
self.ae_net.load_state_dict(model_dict['ae_net_dict'])
def save_results(self, export_json):
"""Save results dict to a JSON-file."""
with open(export_json, 'w') as fp:
json.dump(self.results, fp)